Development and testing of trigger algorithms for the Project 8 experiment

Project 8 is designed to directly measure the neutrino mass using cyclotron radiation emission spectroscopy (CRES). Using cyclotron frequency as a proxy for kinetic energy, CRES can measure the β-decay electron endpoint spectrum for magnetically trapped electrons produced by a gaseous tritium source with high precision. Following the successful demonstration of CRES with waveguides, the upcoming data-taking campaign will demonstrate the first realization of the CRES technique in cylindrical cavities to improve energy resolution by an order of magnitude. CRES events relevant to the Project 8 experiment are rare because only a tiny fraction of events are near the tritium endpoint energy, and the events are short-lived due to the residual-gas scattering of decay electrons within the resonant cavity. This makes full data streaming computationally costly, for which Project 8 is developing an efficient, real-time trigger that will prompt the data acquisition system to record only when a CRES event is present. This talk will report the design and testing of two complementary trigger algorithms: a power-threshold trigger based on CRES signal excess power over background, and a machine learning model based on long short-term memory neural network capable of efficiently detecting patterns of CRES signals buried in noisy data for triggering.